econstor
A Service of
zbw
Make Your Publications Visible.
Leibniz-Informationszentrum
Wirtschaft
Leibniz Information Centre
for Economics
Busso, Matias; Madrigal, Lucia; Pages-Serra, Carmen
Working Paper
Productivity and Resource Misallocation in Latin
America
IDB Working Paper Series, No. IDB-WP-306
Provided in Cooperation with:
Inter-American Development Bank, Washington, DC
Suggested Citation: Busso, Matias; Madrigal, Lucia; Pages-Serra, Carmen (2012) : Productivity
and Resource Misallocation in Latin America, IDB Working Paper Series, No. IDB-WP-306
This Version is available at:
http://hdl.handle.net/10419/88987
Standard-Nutzungsbedingungen:
Terms of use:
Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen
Zwecken und zum Privatgebrauch gespeichert und kopiert werden.
Documents in EconStor may be saved and copied for your
personal and scholarly purposes.
Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle
Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich
machen, vertreiben oder anderweitig nutzen.
You are not to copy documents for public or commercial
purposes, to exhibit the documents publicly, to make them
publicly available on the internet, or to distribute or otherwise
use the documents in public.
Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen
(insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten,
gelten abweichend von diesen Nutzungsbedingungen die in der dort
genannten Lizenz gewährten Nutzungsrechte.
www.econstor.eu
If the documents have been made available under an Open
Content Licence (especially Creative Commons Licences), you
may exercise further usage rights as specified in the indicated
licence.
IDB WORKING PAPER SERIES No. IDB-WP-306
Productivity and
Resource Misallocation
in Latin America
Matías Busso
Lucía Madrigal
Carmen Pagés
April 2012
Inter-American Development Bank
Department of Research and Chief Economist
Productivity and Resource
Misallocation
in Latin America
Matías Busso*
Lucía Madrigal**
Carmen Pagés*
* Inter-American Development Bank and IZA
** Paris School of Economics
Inter-American Development Bank
2012
Cataloging-in-Publication data provided by the
Inter-American Development Bank
Felipe Herrera Library
Busso, Matías.
Productivity and resource misallocation in Latin America / Matías Busso, Lucía Madrigal, Carmen
Pagés.
p. cm. (IDB working paper series ; 306)
Includes bibliographical references.
1. Industrial productivity—Latin America. I. Madrigal, Lucía. II. Pagés, Carmen. III. Inter-American
Development Bank. Research Dept. IV. Title. V. Series.
http://www.iadb.org
Documents published in the IDB working paper series are of the highest academic and editorial quality.
All have been peer reviewed by recognized experts in their field and professionally edited. The
information and opinions presented in these publications are entirely those of the author(s), and no
endorsement by the Inter-American Development Bank, its Board of Executive Directors, or the countries
they represent is expressed or implied.
This paper may be freely reproduced.
Abstract *
Total factor productivity (TFP) in Latin America has not increased relative to the US
since the mid-1970s, and in many countries it has declined. Resource misallocation
can lower aggregate TFP. This paper presents evidence based on firm-level data
from 10 Latin American countries to quantify the heterogeneity of firm productivity
and the degree of resource misallocation within countries. Productivity heterogeneity
and resource misallocation are found to be much larger than in the United States.
Achieving an efficient allocation of resources could boost manufacturing TFP
between 45 percent and 127 percent depending on the countries and years
considered.
JEL classifications: D24, O47, L25, O54
Keywords: Total factor productivity, Firm productivity, Firm heterogeneity,
Distortions, Misallocation costs, Latin America
*
The views expressed here are those of the authors and do not necessarily reflect the opinions of the Inter-American
Development Bank. Corresponding author: Matías Busso ([email protected]).
1
1. Introduction
The gap between Latin America and the developed world has widened dramatically in the last 50
years. In 1955, GDP per capita in Latin America relative to the US was 28 percent. In 2005, it
was 19 percent. Growth accounting exercises indicate that the main reason behind this
divergence has been the low productivity growth experienced by Latin American economies
since the mid-1970s (Hopenhayn and Neumeyer, 2004; Restuccia, 2008; Daude and FernándezArias, 2010).
Aggregate productivity gains are usually thought to be driven mostly by upgrades in the
processes, products or machinery used by firms, affected in turn by investments in human capital
and R&D. In that context, low TFP growth is usually caused by barriers that prevent diffusion
and implementation of new technologies (Parente and Prescott, 2002). However, recent studies
such as Banerjee and Duflo (2005), Restuccia and Rogerson (2008) and Hsieh and Klenow
(2009a) propose an alternative explanation. Low aggregate TFP growth can be also explained by
a number of policy and market failures that determine the selection of firms in the market as well
as the allocation of resources across firms. In the presence of distortions, productive firms are
smaller than they would be in an undistorted economy, therefore lowering aggregate TFP.
In this paper we use micro-data from manufacturing firms in 10 Latin American countries
to measure the extent by which misallocation of resources can explain differences in productivity
between Latin America and the United States. We show that firms’ productivity in Latin
America is very heterogeneous, even within narrowly defined sectors, with a few very
productive firms and many firms of extremely low productivity. We follow the Hsieh and
Klenow (2009b) framework to quantify the potential gains in TFP of reallocating resources
across firms in Latin American countries. We find that reallocating capital and labor to equalize
marginal products in manufacturing would raise aggregate TFP in Latin America between 45 and
127 percent, depending on the countries and years considered.
The results presented in this paper are obtained from the analysis of establishment-level
data produced by countries’ statistical offices. Unfortunately, only a subsample of countries in
Latin America collects establishment-level data and, of those, even a smaller number make data
available to researchers. We are, however, able to use data for 10 Latin American countries:
2
Argentina, Bolivia, Brazil, Chile, Colombia, Ecuador, El Salvador, Mexico, Uruguay and
Venezuela. 1
Data coverage varies across countries but, in general, data are drawn from censuses of the
largest, formally established firms, with a random sample of smaller firms. The datasets are
representative at the national level and span a period of nearly 10 years. In most cases, coverage
is restricted to the manufacturing sector, although in Mexico and Uruguay data coverage extends
beyond manufacturing to other economic sectors. Most of the analysis is restricted to firms of 10
or more employees. When it is available, however, information is also provided on how
measurements change when the smallest firms are added.
2. Firm Productivity and Distortions
Economy-wide productivity growth is typically estimated as the portion of GDP growth that
cannot be explained by either the accumulation of physical and human capital or the growth of
employment. This unexplained part reflects how well countries are able to extract more output
out of a given set of inputs. In recent years, however, a number of new studies are beginning to
look beyond aggregated figures to better understand what drives aggregate productivity
(Barstelman, Haltiwanger, and Scarpetta, 2004; Eslava et al., 2004; Foster, Haltiwanger and
Syverson, 2008; and Syverson, 2004 and 2008). With a few exceptions, this work has been
mostly confined to developed economies.
We measure productivity heterogeneity and distortions across firms in a set of developing
economies in Latin America, following the methodology developed by Hsieh and Klenow
(2009b). In this section we briefly outline their model. Consider a standard model of
monopolistic competition with heterogeneous firms that face distortions in the prices they
observe. These distortions drive wedges between the marginal products of capital and labor
across firms, lowering aggregate TFP.
There is a single final good produced by a representative firm out of a set of goods Ys in a
perfectly competitive final output market with a Cobb-Douglas production technology:
1
The computations in each country were made by the following teams: for Argentina, A. Neumeyer and G.
Sandleris; for Bolivia, C.G. Machicado and J.C. Birbuet; or Brazil, C. Ferraz; for Chile, M. Busso, L. Madrigal,
and C. Pagés; for Colombia, by A. Camacho and E. Conover; for Ecuador, C. Arellano; for El Salvador, J.P. Atal,
M. Busso and C. Cisneros; for Mexico, C. Hsieh, P. Klenow and P. Martínez; for Uruguay, C. Casacuberta and N.
Gandelman; and for Venezuela, L. Kolovitch. In all cases the computations were done using a common program
which is available from the authors upon request.
3
𝑌=�
𝑆
𝑠=1
𝜃
𝑌𝑠 𝑠
with constant returns to scale (so ∑ θs = 1). In turn, each sector output Ys is produced by
combining M differentiated goods Ysi produced by individual firms using a CES technology
𝑌𝑠 =
𝜎−1
𝜎−1 𝜎
�� 𝑌𝑠𝑖 𝜎 �
𝑖=1
𝑀
Each good Ysi is produced with a Cobb-Douglas technology with capital share αs and
productivity Asi
𝛼
1−𝛼𝑠
𝑌𝑠𝑖 = 𝐴𝑠𝑖 𝐾𝑠𝑖𝑠 𝐿𝑠𝑖
where Ysi, Lsi, Ksi, denote output, labor services and capital services. The parameter αs is the
capital share, which is assumed to be constant for all firms within a given industry. As the
elasticity of substitution between plant value-added σ increases, intermediate inputs become
closer to perfect substitutes. At the limit, only the highest-productivity good is produced. The
elasticity of substitution is assumed to be the same for all industries.
Individual firm profits are given by:
π si = (1 − τ Ysi ) PsiYsi − wLsi − (1 + τ Ksi ) RK si
where w, and R denote wages and the rental cost of capital, respectively. There exist two types of
distortions in this economy that affect the decisions of the firms. Output distortions τYsi distort the
output price observed by the firm. These distortions affect both capital and labor. Examples of
these are high transportation costs, bribes/costs that have to be paid in order to operate or
government-issued size restrictions. In turn, capital distortions τKsi change the marginal product
of capital relative to labor. Examples of these are credit constraints and labor market regulations
that due to different credit histories or evasion patterns differ across firms. In the presence of
distortions the marginal revenue products are given by:
MRPK si = R
1 + τ Ksi
1 − τ Ysi
; MRPLsi = w
4
1
1 − τ Ysi
At this point, it is important to distinguish between physical total factor productivity
(TFPQsi), measured by Asi, and total factor revenue productivity (TFPRsi), measured by PsiAsi.. It
can be shown that:
TFPRsi ∝ (MRPK si )
αs
(MRPLsi )
1−α s
αs
R
∝  
 αs 
1−α s
 1 


1 − αs 
(1 + τ Ksi )α
s
1 − τ Ysi
In an economy with no distortions, more inputs should be allocated to firms with the
highest productivity, Asi until TPPRsi is equated across firms within a sector. In such an
economy, there should be no dispersion in the distribution of TFPR. Departures from this
benchmark determine the magnitude of distortions, which are then measured through the
dispersion of TFPR.
It is then possible to relate aggregate TFP to firms’ productivities and firm-level
distortions. Industry TFP can be expressed as a weighted geometric average of firms Asi. Firms
with TFPR smaller than the sector average, that is firms that use more inputs than they would in
an undistorted economy, receive a higher weight. Given the assumed aggregate production
function, aggregate TFP can be expressed as:
θs
θs
S
TFP = ∏ [TFPs ]
s =1
 M  TFPR σ −1  σ −1

s
= ∏  ∑i =1s  Asi
 

s =1
 TFPRsi  

S
In the absence of distortions, aggregate TFP will be higher because resources are
reallocated from less to more productive firms. There will, however, be some dispersion in the
distribution of firms’ productivities. The efficient TFP becomes a geometric average of Asi
S
θs
S
[ ] = ∏ (∑
TFP = ∏ [TFPs ] = ∏ A s
*
s =1
s =1
θs
S
Ms
i =1
s =1
)
θs
σ −1 σ −1
Asi
TFP* can then be used as a benchmark to compute the output cost of deviations from the
efficient allocation of resources. 2 In particular, the gap between the efficient and the actual level
of total factor productivity can be shown to be:
2
αs
R
 1
M  1 + τ Ksi   PsiYsi  
Ms 
1   PsiYsi
TFPR s  ∑ i 1s =
=

⋅
 

∑
i 1
=
s s 
s s
 1 − τ Ysi   PY
 1 − τ Ysi   PY
αS
1 − α S
5
1−α s



θs
 M  A TFPR σ −1  σ −1
TFP

s
= ∏  ∑ s  si

TFP * s =1  i =1  As TFPRsi  


S
Most establishment-level surveys do not record individual, plant or product level prices.
Assuming the model set forth, however, allows estimating physical productivity by means of the
following expression which can be observed in the data.
Asi =
σ
σ −1
(P Y )
Ysi
= α s si si 1−α s
1−α s
K si ( wLsi )
K si ( wLsi )
αs
This result is derived from the fact that product demand is given by Psi = Ysi−1 / σ . In
addition, since workers’ human capital levels are not observable, the plant wage bill is used
instead of labor input as a way of adjusting for differences in human capital across plants.
Finally, αs is measured as one minus the labor share in industry s in the United States. 3 This is a
simple way to control for distortions that could affect the capital share differently in different
countries while the United States is taken as a benchmark of an undistorted economy. The
elasticity of substitution σ is taken to be equal to three across sectors and countries. In the
Appendix, we assess the robustness of the main results to alternative hypothesis about these
parameters.
3. Main Results
Table 1 shows three measures of dispersion of the distribution of log(Asi/ A s): the difference
between the 90th and the 10th percentile, the inter-quartile range and the standard deviation. 4 In
the calculations, sectors are defined at the 4-digit level of disaggregation in the International
Standard Industrial Classification (ISIC). This permits comparison of firms that produce similar
goods.
First note that, while all countries experience some degree of productivity inequality, the
dispersion appears to be greater in Latin America than in the United States. In Colombia and
Venezuela, firms in the 90th percentile of productivity are more than 500 percent more
3
4
As reported by the Manufacturing Industry Database hosted by the NBER.
In the analysis we trim the lowest 1 percent and top 1 percent of the distribution of TFPQ.
6
productive than firms in the 10th percentile, while in the rest of the countries this difference is on
the order of 300 percent, and in the United States the difference is 200 percent. However,
because data coverage varies across countries and dispersion measures are sensitive to the
sample used, direct comparison between results could be misleading. In the United States, data
cover all establishments of one or more employees; in China, establishments with sales above six
hundred thousand dollars per year; and in Latin America, establishments with 10 employees or
more. The bottom panel shows results for all Latin American manufacturing establishments, for
those countries where such figures are available (Mexico and El Salvador) and they also suggest
more dispersion in productivity in Latin America than in the United States.
Table 1. Dispersion of TFPQ (Asi):
Percentile Differences and Standard Deviation (s.d.)
Period
Sample 10 or more workers
Colombia (1982-1998)
Venezuela (1995-2001)
El Salvador (2005)
Chile (1996-2006)
Uruguay (1997-2005)
Bolivia (1988-2001)
Ecuador (1995-2005)
Argentina (1997-2002)
Mexico (1999-2004)
Sample 30 or more workers
Brazil (2000-2005)
El Salvador (2004)
Mexico (1999-2004)
Chile (1996-2006)
Sample: All workers
Mexico (1999-2004)
El Salvador (2004)
China (1998-2005)
US (1977-1997)
90-10
Initial
Final
75-25
Initial
Final
Initial
s.d
4.55
4.42
n.a.
2.65
2.56
3.26
2.78
1.82
3.60
5.51
5.08
3.09
2.92
3.01
2.94
2.86
2.46
3.33
2.52
2.29
n.a.
1.32
1.47
1.57
1.48
1.00
2.05
2.58
2.68
1.54
1.50
1.66
1.71
1.47
1.31
1.73
1.77
1.71
n.a.
1.06
1.00
1.25
1.10
0.71
1.41
2.14
1.91
1.30
1.17
1.20
1.20
1.11
0.94
1.30
3.18
n.a.
3.48
2.60
3.10
2.70
3.17
2.80
0.71
n.a.
1.84
1.27
0.74
1.42
1.72
1.45
1.37
n.a.
1.35
1.03
1.32
1.12
1.23
1.12
4.45
n.a.
2.72
2.22
3.89
2.88
2.44
2.18
2.42
n.a.
1.41
1.22
2.08
1.49
1.28
1.17
1.73
n.a.
1.06
0.85
1.66
1.15
0.95
0.84
Final
This large dispersion implies that, within fairly narrowly defined industries, certain firms
are able to produce much more output than others from the same amount of inputs. This disparity
may be due to extreme variations in the processes and technologies used by firms to produce and
7
compete in the same industry, or it could be related to differences in the human capital or
managerial skill of the managers/owners of firms. Appendix Table A1 shows that differences in
productivity within narrowly defined industries appear to be much higher in non-manufacturing
sectors. While such data are available only for Mexico and Uruguay, in both countries
differences in productivity across firms are much higher in the service sector, particularly in
communication and transportation in Uruguay, and in retail in Mexico.
Using the first-order conditions and assuming that value added does not include any taxes
or subsidies that differentially affect firms within the same industry, we can compute a measure
of the distortions faced by firms:
1 + τ Ksi =
α s wLsi
1 − α s RK si
; 1 + τ Ysi =
σ
wLsi
σ − 1 (1 − α s ) PsiYsi
Table 2 shows the dispersion of the distribution of log(TFPRsi/ TFPR s ), log(τKsi/ τKs ) and
log(τYsi/ τY s ) measured as the difference between the 90th and the 10th percentiles. 5 In an efficient
allocation, marginal returns are equated across firms and therefore the dispersion of marginal
returns and thus of TFPR would be zero. Higher dispersion suggests more misallocation of
resources across plants as a result of heterogeneous policy treatment of firms in the same sector.
Clearly, according to this metric Latin America suffers from a substantial degree of
misallocation. Dispersion in TFPR is lower in the United States, indicating a lower level of
misallocation in this country. Within Latin America, dispersion is higher in Venezuela,
Colombia, Uruguay, and Mexico, all with differences between high and low marginal products
within sectors above 200 percent. Therefore, high levels of dispersion highlight potential gains in
productivity that could be achieved by reallocating factors from firms with low marginal
revenues to those with high marginal revenues. 6
It is worthwhile noting that there exists a higher dispersion in τKs than in τYs suggesting
there are more distortions in input than in product markets. Despite being higher, input markets
5
In the Appendix we provide other dispersion methods for the distributions of the three variables. Also, we trim the
top 1 percent and the bottom 1 percent of the distribution of TFPR. See Appendix Tables A2, A3, and A4.
6
Again, care should be taken in comparing countries with different data coverage. In this case, comparisons for
Mexico and El Salvador, including and excluding the smallest firms, suggest that including these firms may increase
or reduce the estimated degree of misallocation. In Mexico, for example, the dispersion in marginal revenues is 227
percent when all firms are included and only 208 percent for firms of 10 or more employees. In contrast, in El
Salvador, dispersion is 135 percent for all firms and 138 percent for firms of 10 or more employees.
8
distortions were reduced in all countries except in Chile between the initial and the final period.
On the other hand, scale distortions increased everywhere except in Mexico and Bolivia.
Table 2. Dispersion of Distortions (TFPRsi, τYsi and τKsi):
Percentile Difference p90-p10
Period
Sample 10 or more workers
Venezuela (1995-2001)
Colombia (1982-1998)
Uruguay (1997-2005)
Mexico (1999-2004)
Bolivia (1988-2001)
Chile (1996-2006)
Argentina (1997-2002)
Ecuador (1995-2005)
El Salvador (2005)
Sample 30 or more workers
Brazil (2000-2005)
El Salvador (2004)
Mexico (1999-2004)
Chile (1996-2006)
Sample: All workers
El Salvador (2004)
Mexico (2004)
China (1998-2005)
US (1977-1997)
TFPR
Initial
Final
1+τ K
Initial
Final
1-τ Y
Initial
Final
2.60
2.50
2.12
2.33
2.16
1.57
1.04
1.49
n.a.
3.28
2.90
2.47
2.08
2.06
1.77
1.56
1.48
1.39
3.60
n.a.
3.24
3.50
2.65
2.93
2.02
2.81
n.a.
3.17
n.a.
3.16
3.25
2.22
3.23
1.43
2.64
3.31
2.04
0.94
1.00
2.33
1.70
1.37
1.11
1.29
n.a.
2.77
1.41
1.48
2.18
1.53
1.55
2.20
1.34
1.30
1.97
n.a.
2.26
1.55
2.10
1.33
2.04
1.73
2.89
3.57
2.69
3.19
2.58
3.16
2.96
1.91
n.a.
n.a.
1.40
1.98
1.21
n.a.
1.56
n.a.
2.57
1.87
1.04
1.35
2.27
1.59
1.19
n.a.
3.31
n.a.
n.a.
4.11
3.22
n.a.
n.a.
n.a.
2.68
n.a.
n.a.
1.29
2.38
n.a.
n.a.
The large dispersion in both physical productivity and distortions suggests that resources
are not efficiently allocated. We next quantify how costly misallocation is for aggregate
productivity in Latin America. To do so, we compute how much output an economy loses by
allocating the resources inefficiently. In particular, the cost of misallocation is defined as
 TFP *  
 − 1 × 100 .
C = 
TFP
 

Table 3 shows the results. It turns out that by reallocating existing capital and labor
across firms, aggregate productivity in Latin America could increase output between 45 and 127
9
percent depending on the countries, years or samples considered. For most countries the gains
would be around 50-60 percent, with the exception of Mexico, where TFP and GDP per capita
could approximately double if misallocation were to be corrected. These gains consider
reallocation only within four-digit industries. There could be further sizable gains from
reallocating labor and capital across industries.
Table 3. TFP Gains from Equalizing TFPR within Industries
Period
Sample 10 or more workers
Venezuela (1995-2001)
Bolivia (1988-2001)
Uruguay (1997-2005)
Argentina (1997-2002)
Ecuador (1995-2005)
El Salvador (2005)
Chile (1996-2006)
Colombia (1982-1998)
Sample 30 or more workers
Brazil (2000-2005)
El Salvador (2004)
Mexico (1999-2004)
Chile (1996-2006)
Sample: All workers
Mexico (2004)
China (1998-2005)
El Salvador (2004)
US (1977-1997)
Initial
Final
55.2
52.5
61.8
52.2
52.7
n.a.
45.0
48.9
64.7
60.6
60.2
60.0
57.6
60.6
53.8
50.5
49.1
n.a.
140.1
47.5
41.4
55.1
109.5
53.7
127.0
110.5
n.a.
36.1
95.0
86.6
56.7
42.9
Only El Salvador and Mexico have data that cover all firms, which are comparable to the
United States. However, the figures for these countries suggest that the potential gains of
reallocation are larger than in the United States, or, alternatively, that both countries suffer from
more resource misallocation than the United States. Taking the latest figures, improving the
allocation of resources across firms could increase total factor productivity by 95 percent in
Mexico and 56.7 percent in El Salvador, thus helping to close productivity gaps relative to the
United States by a substantial amount. To put these numbers in perspective it is useful to recall
that the typical Latin American country’s TFP is on the order of 55 percent that of the United
10
States, yet Duarte and Restuccia (2010) estimate that across countries, productivity gaps are
lower in manufacturing than across other sectors or in aggregate TFP. This implies that gains
derived from improving the allocation of resources could go a long way towards closing
manufacturing TFP gaps. Moreover, the gains in TFP brought about by an improvement in the
allocation of resources are likely to motivate an increase in the investment rate, as higher
productivity is associated with higher returns of capital and labor.
Appendix Table A5 shows that resource allocation problems appear to be much larger
outside manufacturing, particularly in the service sector. In Uruguay, the potential gains of
reducing misallocation are higher in retail and in the transport and communication sector than
potential gains in manufacturing. In Mexico, the differential in gains across industries is even
larger. While in manufacturing they are on the order of 95 percent, in retail they are 267 percent
and, in the personal and community service sector, 246 percent. De Vries (2009) analyzes the
retail sector of Brazil and finds that the potential gains of reallocating resources towards the most
efficient retailers and finds that they are very large, on the order of 257 percent, enough to move
the productivity of the service sector in Brazil to levels close to those in the United States. These
large gains underscore that a good part of the extremely low productivity in services lies not only
in the low productivity of firms, but also in the poor way resources are allocated across them.
Duarte and Restuccia (2010) point to the lower degree of competition in the service
sectors in relation to manufacturing as one potential reason why across countries there is more
convergence to the world frontier in manufacturing than in the service sector. Services are
generally non-tradable and often heavily protected by a myriad of regulations; moreover,
variables like location play a much more important role in services than in manufacturing.
Extensive misallocation is a symptom of lack of fair competition for resources, as policies,
market failures, or location advantages favor some firms relative to others for reasons other than
their relative efficiency Given the growing importance of the service sector in all economies and
the more rapid growth in productivity of the service sector in the developed world, failure to
improve allocation in this sector could contribute to enlarging the gap in aggregate productivity
relative to higher income countries.
11
4. Firm Size, Distortions and Productivity
We have argued that in economies where workers and capital are poorly allocated across firms,
improving the allocation of resources could provide a boost to productivity comparable to
decades of technological growth. However, making gains effective requires finding out the
sources of misallocation.
In a well-functioning economy, firms that are more productive than their competitors
should win market share over time, hiring more labor and capital and expanding their production.
This implies that firm size (measured either by value added, employment, or assets) should be
strongly positively correlated with firm productivity. 7 Nonetheless, firms do not grow
indefinitely because in order to sell more, they would need to cut prices to a point where they
would make lower profits. The relationship between firm size and productivity breaks down or
becomes weaker if market or government failures favor some firms over others, allowing some
firms to gain market share (size) even if they are less productive, or preventing some firms from
gaining market share even if they are highly productive. This distorts the allocation of resources
across firms, reducing the output that can be attained with existing capital and labor. In this
section we investigate the relation between productivity, distortions and firm size in order to
further understand what is behind the misallocation of resources in Latin America.
We start by establishing the presence of a positive relationship between productivity and
firm size. In table 4 we present simple OLS regressions of log(Asi/ A s) on firm size dummies. 8
Compared to manufacturing firms employing 10–19 workers, manufacturing firms in the 20–49
range are about 50 percent more productive. Productivity more than doubles in firms of more
than 100 workers. In Bolivia, Venezuela, and El Salvador, productivity in the largest firms is
about 150 percent higher than for firms in the 10–19 category. Only in Ecuador does productivity
seem to be completely unrelated to size.
7
The argument is that productivity determines size, with more productive firms growing to be larger, rather than the
other way around: i.e., larger firms become more productive as a result of their size. Yet a positive relationship
between total factor productivity and size can also be driven by economies of scale. This is because most methods of
computing TFP assume constant returns to scale; therefore, increasing returns to scale would wrongly show up as
higher TPF for bigger firms.
8
The data for the selected group of countries presented here also suggest that, in addition to size, productivity
increases with age and with exporting status, although the direction of the causality between exports and
productivity tends to be the reverse: high productivity facilitates exports, and not the other way around. See
Appendix Table A6.
12
Table 4. OLS Regression of Log(Asi/ A s) on Size Dummies
Colombia
Ecuador
Chile
Uruguay
Average
2005
2006
2005
0.439***
-0.019
0.2684***
0.5543**
[0.021]
[0.0486]
[0.0789]
[0.2321]
Size 50-99
1.006***
-0.015
0.3534*
0.6179**
[0.028]
[0.0550]
[0.1920]
[0.2407]
Size 100-249
1.426***
0.073
0.7941***
1.0523***
[0.051]
[0.0782]
[0.1968]
[0.2111]
Size 250-499
1.599***
-0.003
0.8344***
1.3739***
[0.123]
[0.0610]
[0.2345]
[0.2023]
Size 500-999
2.122***
0.012
[0.043]
[0.0557]
Size 1000+
2.338***
[0.053]
Robust standard errors in brackets. *** p<0.01, ** p<0.05, * p<0.1
SIZE
Size 20-49
El Salvador
Bolivia 2000
2004
0.5478***
0.5775**
[0.1025]
[0.2449]
1.0651***
1.3212***
[0.1375]
[0.2930]
1.7283***
2.1632***
[0.1522]
[0.2746]
1.8928***
[0.2379]
Venezuela
2001
0.384
[0.3025]
0.7462***
[0.2639]
1.8532***
[0.2459]
2.1634***
[0.3165]
Brazil
2005
0.473
[0.1626]**
0.849
[0.1554]**
1.324
[0.1479]**
1.905
[0.1903]**
Argentina
-0.121***
(.0126)
0.612***
(.028)
0.682***
(.194)
In order to assess whether distortions affect firms across different size categories we
calculate the percentage of firms that are currently small, medium and large that would decrease
in size if all distortions were eliminated and TFPR was equalized between firms in any given
sector. The results are shown in Table 5. “Small” here means the bottom 25 percent of firms in
the distribution of value added, “large” means the top 25 percent and “medium” the 50 percent in
the middle of the distribution. In general, firms that are currently small are more likely than large
firms to shrink in an efficient allocation of resources. The only exception is Colombia. We
interpret this result as an indication that most distortions affect small firms and that these
distortions go in the direction of favoring larger than efficient sizes for firms that are currently
small firms. In other words, the results suggest that the majority of small firms are not too small
but rather too large given their productivity.
The previous result, combined with the fact that Latin America is a region of very small
firms, at least compared to the United States, helps to explain the productivity gap. When
comparing the subsample of manufacturing establishments with 10 or more workers, the size
distribution of firms across Latin American countries and the United States is quite similar. This
is shown in panel 1 of Table 6. Although in Bolivia, Argentina, Mexico, and El Salvador the
share of small establishments (10–49 workers) is larger than in the United States, in other Latin
American countries this share appears to be at similar or lower levels. It is also interesting to note
that the majority of employment is generated by large firms.
13
Table 5. Percent of Firms That Would Reduce Size in an Efficient Allocation
Venezuela 2001
Brazil 2005
El Salvador 2004
Chile 2006
Ecuador 2005
Uruguay
Bolivia 1997
Colombia
China 2005
US 1997
Small
82.5
81.8
75.8
64.2
52.9
49.4
41.6
31.4
65.6
66.8
Medium
72.5
57.1
47.7
54.3
45.2
41.4
43.0
44.0
55.4
56.6
Large
55.0
40.1
30.6
51.2
35.7
39.0
31.2
69.0
52.4
57.6
Differences across Latin America and the United States appear starker once the smallest
firms are also considered. This is shown in the second panel of Table 6. In Mexico and Bolivia,
91 percent of manufacturing establishments employ fewer than 10 workers. These figures are
lower in Argentina and El Salvador, but still considerably above the percentage of micro firms in
the United States. The share of employment for very small firms is considerably larger in Latin
America than in the United States: about 43 percent of manufacturing employment in Bolivia,
and around 20 percent in Argentina, El Salvador, and Mexico, compared to a mere 4.2 percent in
the United States.
While very few countries have data for all sectors of the economy, when available they
suggest that the percentage of microenterprises is even higher outside manufacturing. In Mexico,
97 percent of establishments in retail and 94 percent in the services sector have fewer than 10
employees, with an average for the whole economy of 95 percent. In retail, 72 percent of
establishments have 2 workers or less (Hsieh and Klenow, 2009b).
14
Table 6. Size Distribution of Establishments and Employment
Panel 1. Establishments of 10 or more Employees
Firm size
[10-19]
[20-49]
[50-99]
[100-249]
[250+]
Firm size
[10-19]
[20-49]
[50-99]
[100-249]
[250+]
Argentina 1993
Est.
Emp.
56.9
17.2
21.6
14.4
15.4
24.3
2.9
10.6
3.3
33.5
Bolivia 1992
Est.
Emp.
51.3
17.7
31.6
24.0
9.9
17.5
7.2
40.8
Chile 2006
Est.
Emp.
26.8
4.0
34.3
10.6
17.3
12.3
12.5
19.8
n.a.
9.2
n.a.
El Salvador 2005
Est.
Emp.
40.8
8.1
29.4
13.4
14.3
15.0
9.5
23.0
6.0
40.5
Mexico 2004
Est.
Emp.
44.3
7.1
28.1
10.4
11.4
9.7
9.3
17.3
6.9
55.5
Argentina 1994
Est.
Emp.
84.0
22
12.9
25
2.5
19
0.8
35
0.2
18
Bolivia 1992
Est.
Emp.
91.7
43.6
4.2
10.0
2.6
13.6
0.8
9.8
0.6
23.0
Colombia 1998
Est.
Emp.
28.7
4.56
31.6
11.07
18.2
14.49
13.9
24.48
53.3
Uruguay 2005
Est.
Emp.
15.4
2.1
34.9
11.2
23.4
16.0
17.5
25.2
8.8
45.5
7.6
45.4
Venezuela 2001
Est.
Emp.
16.5
1.5
25.3
5.2
15.9
7.0
25.1
26.9
17.3
59.4
Ecuador 2005
Est.
Emp.
30.3
3.9
31.2
8.9
16.0
10.2
13.4
19.8
9.1
57.3
US 2003
Est.
Emp.
31.9
5.0
32.4
11.5
16.2
12.9
12.8
22.2
6.8
48.4
Panel 2. All establishments
Firm Size
[1-9]
[10-19]
[20-49]
[50-99]
[100+]
El Salvador 2005
Est.
Emp.
82.0
17.7
8.3
6.2
3.9
6.2
2.8
10.2
2.9
59.7
Mexico 2004
Est.
Emp.
90.5
22.7
4.2
5.5
2.7
8
1.1
7.5
1.6
56.3
US 2003
Est.
Emp.
54.5
4.2
14.5
4.8
14.7
11
7.4
12.3
8.9
67.7
This higher prevalence of smaller firms is underestimated in establishment-level data,
even if it comes from a census, as in this case, since it takes into account only establishments
with a fixed location. Itinerant businesses or street vendors are not usually included in census
data. In Mexico, establishments covered by the economic census account for only 40 percent of
the labor force. Another 26 percent are employed in sectors such as agriculture or government
that are not surveyed. This leaves a very sizeable 13.6 million workers (33.5 percent of the labor
force) unaccounted for. Data from employment surveys (INEGI, 2003) indicate that these
workers work in mobile locations without a fixed establishment, of which 5 million work on
their own, and 6 million in firms with fewer than five workers. By sector, the percentage of
workers without a fixed establishment not accounted for in the census is 8.4 percent for
manufacturing, 17 percent for retail and commerce, and 95 percent for non-financial services. 9
9
Calculations from INEGI (2002).
15
5. Discussion: Misallocation and Policies
What are the potential sources of misallocation? One of the most obvious culprits in resource
misallocation is the financial market. Financial markets in Latin America are underdeveloped
and leave many firms underserved. If financial institutions are unable or unwilling to provide
credit to firms that are highly productive but that have no credit history or insufficient
guarantees, then these firms cannot expand as far as their ideas/projects could take them if
markets worked properly. In an economy where good firms are credit-constrained, transferring
additional resources to these firms can yield very high returns. Resource misallocation can also
occur if directed credit provides cheap credit to inefficient firms, thereby allowing inefficient
firms to expand. 10
The second suspect in resource misallocation is the tax collection system.
The
combination of high taxes and poor enforcement creates strong incentives for tax evasion in
Latin America. Moreover, in many countries, tax authorities searching for ways to improve the
efficiency of tax collection focus their enforcement activity on the largest and most productive
firms, virtually ignoring tax collection from micro, small, and medium enterprises. Since the sum
of taxes and regulation compliance may be high in Latin America (particularly in the highestincome countries), noncompliance with taxes is equivalent to a substantial subsidy to
noncompliant, less productive firms. Since large firms tend to be more productive than smaller
firms, selective noncompliance amounts to a potentially large subsidy to less productive, smaller
firms, thereby artificially increasing their size and weight in the economy while constraining the
size of larger, more productive firms.
A third suspect is the poor enforcement and incomplete coverage of social security
systems. In addition to leaving many workers unprotected, uneven enforcement of those who are
covered can also have negative effects on resource allocation and productivity. By evading taxes,
some firms can save on a number of costs associated with taxes and regulatory mandates, and
therefore compete on unfair terms with more productive firms. To the extent that firm evasion is
triggered by a deliberate attempt to compete with more productive firms, resources may be
diverted from the best firms, promoting instead the expansion and/or survival of less efficient
ones. By evading labor and social security regulations, less productive firms can divert resources
away from larger, more productive firms. These effects can be magnified by the fact that
10
Another reason a firm may be inefficiently small is that it exerts some form of monopoly power.
16
governments increasingly provide some benefits (health, pension) free of charge to workers
conditional on their not being affiliated with social security (Levy, 2008). While the benefits of
such programs for the underserved population may be large, the adverse effects on productivity
may be also sizeable if they simply fuel the fire and help many workers—not satisfied with the
value of services offered by social security—to switch toward self-employment or very small
firms where they can avoid paying social security contributions and still get some of the benefits
free of charge, provided that they remain informal.
These three possible drivers of misallocation suggest predictions regarding the
relationship between misallocation and firm size that can help identify the source of
misallocation. If misallocation is due to financial market failures, this would be reflected in the
presence of many small firms that have difficulty growing—even though they have good
projects—because they cannot secure access to credit. For these firms, the returns of additional
capital would be very large—much larger than for firms whose demands for funding have been
met by the capital market and therefore do not have any high return projects left to fund. If
credit markets are the problem, then on average returns to additional factors would be higher in
small firms than in larger ones. On the other hand, if distortions are due to unequal enforcement
of taxes, social security contributions, or labor regulations, then the returns of additional capital
and labor would be expected to be lower in smaller firms. This is because noncompliant firms
are generally small, and tax evasion works as a subsidy that helps them expand beyond what they
would have had they paid taxes, lowering the marginal returns of factors relative to compliant
firms.
What does the relationship between marginal products of factors and firm size reveal
about the origins of misallocation? We explore this in Table 7. This relationship varies across
countries, but more often than not returns are increasing in relation to firm size. In some
countries, such as Colombia, El Salvador, and Mexico, the marginal revenue product of an extra
unit of resources tends to be larger in medium and large firms than in the smallest ones. In these
countries, evidence suggests that providing extra resources to medium or large firms would yield
higher returns than providing resources to smaller ones. The implication is that in these
countries, most small firms are not too small, but rather too large relative to what they should be
in an efficient allocation. In contrast, medium and large firms appear to be too small relative to
what they would be if resources were assigned following relative productivities. In these
17
countries it is difficult to argue that the main source of such distortions are capital market
constraints, unless it can be shown that medium or large firms are the most constrained by lack
of financial access. Instead, it might well be that small firms are credit constrained but
compensate for these higher costs—or for greater difficulty in accessing credit—by not paying
taxes and circumventing regulations. This latter effect seems to dominate. In this set of countries,
tax evasion and informality concentrated in the smallest firms are very plausible sources of
misallocation.
Table 7. OLS Regression of Log(TFPRsi/ TFPR s ) on Size Dummies
Colombia
Ecuador
Uruguay
Chile 2006
Average
2005
2005
0.010
-0.0930*
-0.0801**
0.091
[0.015]
[0.0519]
[0.0376]
[0.1736]
Size 50-99
0.099***
-0.2033*** -0.2241***
-0.3159*
[0.019]
[0.0587]
[0.0754]
[0.1806]
Size 100-249
0.169***
-0.021
-0.0772
-0.179
[0.030]
[0.0688]
[0.0508]
[0.1878]
Size 250-499
0.085
-0.2797*** -0.5109***
-0.3216*
[0.061]
[0.0800]
[0.1160]
[0.1759]
Size 500-999
0.212***
[0.026]
Size 1000+
0.237***
[0.046]
Robust standard errors in brackets *** p<0.01, ** p<0.05, * p<0.1
SIZE
Size 20-49
El Salvador
Venezuela
Bolivia 2000
Brazil 2005
2004
2001
0.1227*
0.053
-0.136
0.141
[0.0667]
[0.1798]
[0.2177]
[0.0797]
0.138
-0.208
-0.291
0.248
[0.0875]
[0.2041]
[0.1968]
[0.0939]**
0.3390***
-0.029
0.126
0.303
[0.1009]
[0.1789]
[0.1464]
[0.0978]**
0.188
0.075
0.408
[0.1458]
[0.2139]
[0.1472]**
Argentina
-0.121***
(0.0126)
0.612***
(0.028)
0.682***
(0.194)
Patterns differ in some countries, however. Credit market constraints, for instance, seem
a more likely source of distortions in Uruguay and Chile. In these two countries, the returns to an
extra unit of capital and labor tend to decline with firm size, indicating that the smallest firms
tend to be size constrained, while the largest firms appear subsidized, given their productivity.
The lower level of evasion and higher level of formality in Uruguay and Chile may also explain
why in these two countries small firms are relatively more size constrained, as they cannot easily
compensate for low access to credit with tax and social security evasion. However, tax evasion
favoring the largest firms could also explain these patterns. In Chile, larger firms evade taxes
more than smaller ones (see Busso, Madrigal, and Pagés, 2010). In addition, in Chile, larger
manufacturing firms appear to receive more state subsidies than smaller firms, leading again to
higher implicit subsidies for larger firms.
The evidence presented provides some interesting clues as to the likely sources of
misallocation in Latin America. Contrary to the conventional wisdom, there is not much
18
evidence for the hypothesis that very small firms are too small, or that they are size constrained.
Only in Chile and Uruguay are marginal products of capital clearly declining in size, as this
hypothesis would imply, and in Chile tax avoidance of larger firms can also account for this
pattern. As indicated, even if small firms suffer from capital access constraints, other factors,
such as their partial or total noncompliance with taxes and social security mandates, provide
them with an implicit subsidy that allows them to be larger than the size that would be warranted
by their productivity.
6. Conclusion
In this paper we apply the methodology developed by Hsieh and Klenow (2009a) to assess the
extent of heterogeneity of firm productivity and distortions within narrowly defined sectors. We
found that dispersion in these two measures is much larger than in the United States, suggesting
the possibility of gains from moving to an efficient allocation of resources in manufacturing. We
show that indeed those gains are on the order of 60 percent. These large disparities in
productivity and substantial resource misallocation open important avenues for productivity
growth. While the gains from improving resource allocation and the mix of firms would provide
only temporary sources of growth, they could provide a huge leap forward similar to what the
region enjoyed during the period of rapid urbanization and structural transformation during the
1950s and 1960s. This transformation would require reforms aimed at reducing the distortions
created by differences in tax codes and uneven enforcement of taxes and regulations, improving
social insurance policies, improving the functioning of capital markets, and stimulating
competition, particularly in service sectors.
19
References
Arellano, C. 2009. “Productivity Levels, Dispersion and Growth in Ecuador.” Washington, DC,
United States: Inter-American Development Bank, Research Department. Unpublished
document.
Atal, J., M. Busso and C. Cisneros. 2009. “Productivity and Misallocation: The Case of El
Salvador.” Washington, DC, United States: Inter-American Development Bank, Research
Department. Unpublished document.
Banerjee, A., and E. Duflo. 2005. “Growth Theory through the Lens of Development
Economics.” In: P. Aghion and P. Durlauf, editors. Handbook of Economic Growth.
Volume 1a. Amsterdam, The Netherlands: Elsevier.
Barstelman, E., J. Haltiwanger, and S. Scarpetta, 2004. “Microeconomic Evidence of Creative
Destruction in Industrial and Developing Countries.” IZA Discussion Paper 1374. Bonn,
Germany: Institute for the Study of Labor (IZA).
BLS (U.S. Bureau of Labor Statistics). 2005. Business Employment Dynamics. Available at:
http://www.bls.gov/bdm/; and http://www.bls.gov/news.release/pdf/cewfs.pdf.
Busso M., L. Madrigal and C. Pagés. 2010. “Reported Tax Evasion and Resource Misallocation
in Chile.” Washington, DC, United States: Inter-American Development Bank, Research
Department. Unpublished document.
Camacho, A., and E. Conover. 2009. “Misallocation and Manufacturing TFP in Colombia.”
Washington, DC, United States: Inter-American Development Bank, Research
Department. Unpublished document.
Carpio, S., and C. Pagés. 2009. “Informality, Misallocation and Productivity in Brazil.”
Washington, DC, United States: Inter-American Development Bank, Research
Department. Unpublished document.
Casacuberta, C., and N. Gandelman. 2009. “Productivity, Exit and Crisis in Uruguayan
Manufacturing and Service Sectors.” Washington, DC, United States: Inter-American
Development Bank, Research Department. Unpublished document.
Cedlas (Centro de Estudios Distributivos Laborales y Sociales) and World Bank. 2009. SocioEconomic Database for Latin America and the Caribbean (SEDLAC). Available at:
http://www.depeco.econo.unlp.edu.ar/sedlac/eng/index.php/.
20
Daude, C., and E. Fernández-Arias. 2010. “On the Role of Aggregate Productivity and Factor
Accumulation in Economic Development in Latin America and the Caribbean.” IDB
Working Paper IDB-WP-131. Washington, DC, United States: Inter-American
Development Bank.
De Vries, Gaaitzen. 2009. “Productivity in a Distorted Market: The Case of Brazil’s Retail
Sector.” Memorandum GD-112. Groningen, The Netherlands: University of Groningen,
Groningen Growth and Development Centre.
Duarte, M., and D. Restuccia. 2010. “The Role of the Structural Transformation in Aggregate
Productivity.” Quarterly Journal of Economics 125(1): 129-173.
Eslava, M. et al. 2004. “The Effect of Structural Reforms on Productivity and Profitability
Enhancing Reallocation: Evidence from Colombia.” Journal of Development Economics
75: 333-371.
Foster L., J. Haltiwanger and C. Krizan. 2001. “Aggregate Productivity Growth: Lessons from
Microeconomic Evidence.” In: C.R. Hulten, E.R. Dean and M.J. Harper, editors. New
Developments in Productivity Analysis. NBER Book Series Studies in Income and
Wealth. Chicago, United States: University of Chicago Press.
Foster L., J. Haltiwanger and C. Syverson. 2008. “Reallocation, Firm Turnover, and Efficiency:
Selection on Productivity or Profitability?” American Economic Review 98(1): 394-425.
Hopenhayn, H., and P.A. Neumeyer. 2004. “Latin America in the Twentieth Century:
Stagnation, then Collapse.” Department of Economics Working Papers 028. Buenos
Aires, Argentina: Universidad Torcuato Di Tella.
Hsieh, C., and P. Klenow. 2009a. “Misallocation and Manufacturing TFP in China and India.”
Quarterly Journal of Economics 124(4): 1403-1448.
Hsieh, C., and P. Klenow. 2009b. “Misallocation in Mexico.” Washington, DC, United States:
Inter-American Development Bank, Research Department. Unpublished document.
INE (Instituto Nacional de Estadística, Venezuela). 2001. Annual Industrial Survey. Caracas,
Venezuela: INE.
INEGI (Instituto Nacional de Estadística y Geografía, Mexico). 2002. Micro-firms National
Survey. Mexico. Available at:
http://www.inegi.org.mx/est/contenidos/espanol/sistemas/enamin/enamin_2002/default.asp.
21
INEGI. 2003. National Survey of Employment and Occupation. Mexico. Available at:
http://www.inegi.org.mx/inegi/default.aspx?s=est&c=10658
INEGI. 2004. Economic Census Questionnaire. Mexico. Available at:
http://www.inegi.org.mx/inegi/default.aspx?c=9745&s=est.
INEGI.
2005.
Results
of
the
Economic
Census.
México.
Available
at:
http://www.inegi.gob.mx/est/contenidos/espanol/proyectos/censos/ce2004/pdfs/resultados_grals.pdf.
Jeong, H., and R. Townsend. 2007. “Sources of TFP Growth: Occupational Choice and Financial
Deepening.” Economic Theory 32: 179–221.
Levy, S. 2008. Good Intentions, Bad Outcomes: Social Policy, Informality, and Economic
Growth in Mexico. Washington, DC, United States: Brookings Institution Press.
Machicado, C.G., and J.C. Birbuet. 2009. “Understanding Productivity Levels, Dispersion and
Growth in the Leather Shoe Industry: Effects of Size and Informality.” Washington, DC,
United States: Inter-American Development Bank, Research Department. Unpublished
document.
Neumeyer, A., and G. Sandleris. 2009. “Productivity and Resource Misallocation in the
Argentine Manufacturing Sector, 1997–2002.” Washington, DC, United States: InterAmerican Development Bank, Research Department. Unpublished document.
Parente, S.L., and E.C. Prescott. 2002. Barriers to Riches. Cambridge, United States: MIT Press.
Restuccia, D. 2008. “The Latin American Development Problem.” Working Paper 318. Toronto,
Canada: University of Toronto, Economics Department.
Restuccia, D., and R. Rogerson. 2008. “Policy Distortions and Aggregate Productivity with
Heterogeneous Establishments.” Review of Economic Dynamics 11(4): 707-720.
Syverson, C. 2004. “Market Structure and Productivity: A Concrete Example.” Journal of
Political Economy 112(6): 1181–1222.
Syverson, C. 2008. “Markets: Ready-Mixed Concrete.” Journal of Economic Perspectives 22(1):
217–33.
World Bank. 2006. Enterprise Survey. Available at: http://espanol.enterprisesurveys.org/Default.aspx.
22
Appendix
1. Robustness
Table A1 compares the hypothetical TFP gains in our “benchmark” assuming an elasticity of
substitution within industries of 3, with TFP gains calculated with a higher elasticity of 5. As
expected, in most countries, TFP gains increase. When the elasticity of substitution is higher the
process of reallocation of resources within industries is slower, which explains potential higher
TFP gains. Hsieh and Klenow (2009a) found that, changing the same assumption, China’s
hypothetical TFP gain in 2005 climbs from 87 percent under σ = 3 to 184 percent with σ = 5, and
India’s in 1994 from 128 percent to 230 percent. Results in Latin American countries increase
but not so dramatically, showing that gains are indeed sensitive to the elasticity of substitution.
Table A1. TFP Gains, Robustness
Benchmark
Initial
Sample 10 or more workers
Bolivia (1988-2001)
52.5
El Salvador (2005)
n.a.
Uruguay (1997-2005)
61.8
Argentina (1997-2002)
52.2
Ecuador (1995-2005)
52.7
Chile (1996-2006)
45.0
Colombia (1982-1998)
48.9
Mexico (1999-2004)
Sample 30 or more workers
Brazil (2000-2005)
49.1
sigma=5
local shares
Final
Initial
Final
Initial
Final
60.6
48.2
60.2
60.0
57.6
53.8
50.5
98.6
n.a.
74.4
74.0
36.0
40.3
75.5
99.8
77.8
77.8
80.5
59.0
50.7
78.6
55.4
n.a.
51.8
46.9
52.0
71.6
83.4
48.6
107.0
49.0
54.6
55.0
69.2
73.3
41.4
44.3
52.1
57.7
61.5
When estimating TPF gains in order to control for distortions that could affect the capital
share differently in different countries, we use αs, which is measured as one minus the labor
share in industry s in the United States, which is assumed to be undistorted. In Table A1, we
present TFP gains using the actual distribution of capital in each country. The results are
somewhat mixed: in some countries TFP gains increase with local shares, and in some of the
others the gains are roughly the same or decrease.
We also calculated the dispersion of TFPQ with an elasticity of substitution within
industries of 5 and using local capital shares, as is shown in Table A2. The dispersion
23
consistently diminishes when assuming intermediate inputs become closer to perfect substitutes.
When assuming capital shares of the country instead of US shares, in most countries the results
are unchanged with respect to the benchmark estimation.
Table A2. Dispersion of TFPQ, Robustness
Benchmark
Period
Sample 10 or more workers
Mexico (1999-2004)
El Salvador (2005)
Uruguay (1997-2005)
Chile (1996-2006)
Ecuador (1995-2005)
Argentina (1997-2002)
Sample 30 or more workers
Brazil (2000-2005)
sigma=5
local shares
90-10
Initial
Final
90-10
Initial
Final
90-10
Initial
Final
3.60
n.a.
2.56
2.65
2.78
1.82
3.33
2.46
3.01
2.92
2.86
2.46
n.a.
n.a.
2.42
2.20
n.a.
1.50
n.a.
1.79
2.87
2.51
n.a.
2.00
n.a.
n.a.
2.30
3.04
2.87
1.84
n.a.
2.82
2.73
3.04
2.78
2.44
3.18
3.10
2.89
2.58
3.30
3.35
In Table A3 we present the results for TFPR and the capital and output wedges. Overall,
the results are robust to changes in the assumptions of the model. TFPR dispersion remains more
or less the same, and changes over time are also as previously found when assuming an elasticity
of substitution of 5. This results hold for capital and output wedges as it is shown in Table A3.
Likewise, when using local shares in the estimation of the TFPR, 1+τK , and 1-τY the results are
broadly unchanged.
24
Table A3. Dispersion of Distortions, Robustness
Measured as p90-p10
TFPR
Period
Sample 10 or more workers
Uruguay (1997-2005)
Mexico (1999-2004)
Chile (1996-2006)
Argentina (1997-2002)
Ecuador (1995-2005)
El Salvador (2005)
Sample 30 or more workers
Brazil (2000-2005)
1+τ K
Period
Sample 10 or more workers
El Salvador (2005)
Mexico (1999-2004)
Chile (1996-2006)
Uruguay (1997-2005)
Ecuador (1995-2005)
Argentina (1997-2002)
Sample 30 or more workers
Brazil (2000-2005)
1-τ Y
Period
Sample 10 or more workers
Argentina (1997-2002)
Mexico (1999-2004)
Chile (1996-2006)
Uruguay (1997-2005)
Ecuador (1995-2005)
El Salvador (2005)
Sample 30 or more workers
Brazil (2000-2005)
Benchmark
sigma=5
local shares
Initial
Final
Initial
Final
Initial
Final
2.12
2.33
1.57
1.04
1.49
n.a.
2.47
2.08
1.77
1.56
1.48
1.29
2.12
n.a.
1.64
1.19
n.a.
n.a.
2.50
n.a.
1.93
1.60
n.a.
1.29
1.69
n.a.
2.26
1.24
1.49
n.a.
2.05
n.a.
2.28
1.59
1.47
2.00
1.97
2.10
Benchmark
sigma=5
local shares
Initial
Final
Initial
Final
Initial
Final
n.a.
3.50
2.93
3.24
2.81
2.02
2.43
3.25
3.23
3.16
2.64
1.43
n.a.
n.a.
3.28
3.26
n.a.
2.01
2.43
n.a.
3.34
3.15
n.a.
2.23
n.a.
n.a.
3.26
3.28
1.30
2.01
2.42
n.a.
3.45
3.17
1.23
2.23
2.89
3.19
Benchmark
sigma=5
local shares
Initial
Final
Initial
Final
Initial
Final
1.11
2.33
1.37
1.00
1.29
n.a.
2.20
2.18
1.55
1.48
1.34
1.08
1.09
n.a.
1.48
1.00
n.a.
n.a.
1.50
n.a.
1.73
1.58
n.a.
1.08
1.09
n.a.
1.50
1.01
2.85
n.a.
1.50
n.a.
1.75
1.49
2.71
1.10
1.91
1.98
n.a.
n.a.
n.a.
n.a.
Note: Initial year in Ecuador for other than benchmark is 1997.
25
2. Dispersion and Correlation Measures
A4. Dispersion of TFPR
Period
Initial
90-10
Final
Initial
2.60
2.50
2.12
2.33
2.16
1.57
1.04
1.49
n.a.
3.28
2.90
2.47
2.08
2.06
1.77
1.56
1.48
1.39
1.97
n.a.
2.26
1.55
n.a.
2.57
1.87
1.04
75-25
s.d
Final
Initial
Final
1.27
1.21
1.09
1.27
1.06
0.73
0.68
0.73
n.a.
1.77
1.28
1.24
1.09
0.97
0.86
0.87
0.74
0.74
1.16
1.00
0.83
0.93
0.91
0.66
0.48
0.63
n.a.
1.28
1.21
0.97
0.82
0.88
0.72
0.62
0.62
0.64
2.10
1.33
2.04
1.73
0.40
n.a.
1.16
0.76
0.45
0.66
1.05
0.86
0.89
n.a.
0.90
0.64
0.90
0.64
0.81
0.69
1.35
2.27
1.59
1.19
n.a.
1.33
0.97
0.46
0.69
1.18
0.82
0.53
n.a.
1.02
0.74
0.45
0.58
0.98
0.63
0.49
Final
Initial
Sample 10 or more workers
Venezuela (1995-2001)
Colombia (1982-1998)
Uruguay (1997-2005)
Mexico (1999-2004)
Bolivia (1988-2001)
Chile (1996-2006)
Argentina (1997-2002)
Ecuador (1995-2005)
El Salvador (2005)
Sample 30 or more workers
Brazil (2000-2005)
El Salvador (2004)
Mexico (1999-2004)
Chile (1996-2006)
Sample: All workers
El Salvador (2004)
Mexico (2004)
China (1998-2005)
US (1977-1997)
A5. Dispersion of Capital Wedge
Period
Initial
90-10
75-25
s.d
Final
Initial
Final
2.93
n.a.
3.50
3.24
3.60
2.81
2.02
2.65
3.23
3.31
3.25
3.16
3.17
2.64
1.43
2.22
1.50
n.a.
1.84
1.64
1.78
1.37
1.06
1.27
1.68
1.54
1.75
1.61
1.51
1.29
0.72
1.02
1.31
n.a.
1.35
1.23
1.46
1.16
0.79
1.07
1.37
1.38
1.27
1.31
1.34
1.13
0.86
1.00
2.89
n.a.
3.57
2.69
3.19
2.58
3.16
2.96
0.75
n.a.
1.91
1.38
0.85
1.18
1.64
1.57
1.31
n.a.
1.39
1.15
1.43
1.23
1.23
1.27
3.31
n.a.
3.22
4.11
1.75
n.a.
1.78
2.22
1.29
n.a.
1.36
1.65
Sample 10 or more workers
Chile (1996-2006)
El Salvador (2005)
Mexico (1999-2004)
Uruguay (1997-2005)
Venezuela (1995-2001)
Ecuador (1995-2005)
Argentina (1997-2002)
Bolivia (1988-2001)
Sample 30 or more workers
Brazil (2000-2005)
El Salvador (2004)
Mexico (1999-2004)
Chile (1996-2006)
Sample: All workers
Mexico (1999-2004)
El Salvador (2004)
26
A6. Dispersion of Output Wedge
Period
Initial
90-10
Final
Initial
2.33
1.11
2.04
1.37
1.70
0.94
1.00
1.29
n.a.
2.18
2.20
2.77
1.55
1.53
1.41
1.48
1.34
1.30
1.91
n.a.
1.40
2.68
n.a.
75-25
s.d
Final
Initial
Final
1.24
0.55
1.06
0.68
0.85
0.45
0.46
0.62
n.a.
1.13
1.22
1.26
0.73
0.67
0.60
0.72
0.61
0.63
0.91
0.42
0.85
0.61
0.76
0.58
0.47
0.58
n.a.
0.86
0.57
1.13
0.65
0.62
1.25
0.61
0.59
0.60
1.98
1.21
1.56
0.49
n.a.
0.69
0.50
0.60
0.75
0.87
n.a.
0.60
0.88
0.59
0.66
2.38
1.29
1.40
n.a.
1.24
0.63
1.06
n.a.
1.02
0.56
Sample 10 or more workers
Mexico (1999-2004)
Argentina (1997-2002)
Venezuela (1995-2001)
Chile (1996-2006)
Bolivia (1988-2001)
Colombia (1982-1998)
Uruguay (1997-2005)
Ecuador (1995-2005)
El Salvador (2005)
Sample 30 or more workers
Brazil (2000-2005)
El Salvador (2004)
Chile (1996-2006)
Sample: All workers
Mexico (1999-2004)
El Salvador (2004)
A7. Correlates of TFPQ
AGE
Age 6 - 10
Colombia
Average
Dependent variable: log(TFPQ/TFPQ_bar)
Chile
Uruguay El Salvador
Venezuela
2006
2005
2004
Bolivia 2000
2001
0.191***
[0.040]
0.414***
[0.043]
74392
0.020
-0.624
[0.4716]
-0.107
[0.2365]
3777
0.022
1.6152***
[0.4332]
0.000
[0.0000]
432
0.020
n.a.
1.088***
[0.028]
Robust standard errors in brackets
*** p<0.01, ** p<0.05, * p<0.1
0.1166**
[0.0460]
0.000
[0.0000]
0.265
[6.17]**
Age 10 and more
Observations
R-squared
EXPORTER
Exporter
n.a.
27
-0.611
[0.3865]
0.052
[0.3162]
0.004
[0.0415]
0.140
[0.3792]
0.4806*
[0.2411]
468
0.009
0.101
[0.0942]
Brazil
2005
0.510
[0.0484]**
Argentina
-0.119***
[-0.039]
-0.062*
[0.036]
22397
0.030
0.017
[0.0249]
0.295***
[0.032]